Relationships among measures of longest utterances, MLU, age, and number of utterances in child language samples

Abstract

The length of longest utterances (LLU) may offer a useful view of the upper boundaries of a child's early expressive grammar. In this study, the properties of four measures of LLU were examined in terms of their relationships to mean length of utterance (MLU) and to language sample size over the course of early development. The four measures of LLU were (1) mean length of the longest three utterances (LLU-3), (2) mean length of the longest 10% of utterances (LLU-10%), (3) MLUs over the median utterance length (LLU-M), and (4) MLUs longer than one morpheme (LLU > 1). Language samples of various sizes were created by randomly selecting subsets of 50, 100, 200, and 300 utterances from Brown's samples of Adam, Eve, and Sarah. The four calculation methods were compared in each of the sample sizes, over the course of development for each child. The results indicated that LLU-M and LLU > 1 were both close to MLU, although age differences were noted. Increases in LLU tended to outpace MLU over development in three of the four calculation methods. LLU was positively related to language sample size for one of the calculation methods, LLU-3. Overall, LLU was found to be a fairly close parallel of MLU and a good predictor of future MLU values. It was concluded that three basic properties of LLU should be taken into consideration: (1) calculation method affects LLU outcome, (2) LLU gradually increases relative to MLU throughout early development, and (3) LLU is affected by interaction of language sample size with calculation method.